Table 6 ROC AUC scores comparison (non-stratified) between KGE-based models with and without semantic types information (ST) added to the training set
From: Dyport: dynamic importance-based biomedical hypothesis generation benchmarking technique
| Â | ComplEx | DistMult | HolE | TransE |
|---|---|---|---|---|
No ST | 0.564 | 0.680 | 0.595 | 0.549 |
ST | 0.693 | 0.697 | 0.718 | 0.678 |